[Ironruby-core] Status of rubygems unit tests

Jimmy Schementi Jimmy.Schementi at microsoft.com
Fri Apr 3 17:36:35 EDT 2009

Thanks for the update. I can add this to the runs of http://ironruby.info so we have up-to-date stats about RubyGems.

From: ironruby-core-bounces at rubyforge.org [mailto:ironruby-core-bounces at rubyforge.org] On Behalf Of Shri Borde
Sent: Friday, April 03, 2009 2:15 PM
To: ironruby-core at rubyforge.org
Subject: Re: [Ironruby-core] Status of rubygems unit tests

Here are the latest results:

728 tests, 2021 assertions, 26 failures, 35 errors, 4 skips

Latests results will be available at http://www.ironruby.net/Documentation/How_do_I_install_additional_gems%3f. The total number of tests increased because we moved from RubyGems version 0.9 something to the trunk of 1.3.

Jirapong has been tagging with me to help push on this which has been very useful.

 The issues we have fixed include:

*         GzipWriter.finish should close the GzipWriter object but not the underlying IO object

*         File.chmod was a nop before

*         If a gem supports multiple platforms, the output is non-deterministic since the code iterates a Hash. The keys of a hash are not required to be in any order. Jirapong is going to submit a patch to RubyGems.

*         Jirapong is working on a fix where File.open(File::EXCL) should not keep the file locked. This could fix 8 errors.


From: ironruby-core-bounces at rubyforge.org [mailto:ironruby-core-bounces at rubyforge.org] On Behalf Of Shri Borde
Sent: Monday, March 16, 2009 11:10 PM
To: 'ironruby-core at rubyforge.org'
Subject: [Ironruby-core] Status of rubygems unit tests

I have been working on making  the rubygem unit tests (from git://github.com/vvs/rubygems.git) pass with IronRuby. When I started, the results looked like this.

707 tests, 1170 assertions, 9 failures, 352 errors, 1 skips

Half the tests (9+352/707) were failing (or half were passing depending on how you want to look at it). The main culprits were Zlib::GzipWriter, YAML.quick_emit and Zlib::Deflate. After fixing these issues, the results look like this:

707 tests, 1881 assertions, 40 failures, 53 errors, 4 skips

It's a good improvement. Errors (where the tests cannot even run properly) account for 8% (53/707), and failures (functionality defects) can now show through - in 5% (40/707) of the tests, for a pass rate of 87%. The errors are because of the following exceptions in descending order of occurrence. The number at the start is the count of how often the error happened.

9: dump format error(9)
8: ERROR: Failed to build gem native extension.
5: dump format error(8)
5: marshal data too short
4: uninitialized constant OpenSSL::X509
2: uninitialized constant OpenSSL::PKey
2: trying to seek on a non-existent stream?
2: ERROR: Failed to build gem native extension.
2: SSL is not installed on this system
2: SocketError: An established connection was aborted by the software in your host machine

I will keep pushing on the tests. The failures will be a bit more work to narrow down as some expectation fails, possibly much after a bug introduced a bad result in the system. If anyone wants to help dig into the failures, let me know, and I can help you get set up.


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://rubyforge.org/pipermail/ironruby-core/attachments/20090403/374cdfda/attachment-0001.html>

More information about the Ironruby-core mailing list